Hi, 

>>I am also confused that people seem to think that INFOs will start 
>>flying around for no reason, without the sender having any 
>>clue about whether the receiver will understand them or not. As we
have 
>>discussed, there ARE mechanisms in SIP to negotiate what content types

>>etc entities support. Yes, MAYBE we need to extend that (e.g.
>>being able to indicate which methods can be used to carry a 
>>specific content type), if what we have is not enough. But, to me 
>>that is not the same thing has having "tons of problem".
> 
>There is a big difference between, "I can support this 
>format", and "I am interested in having you send me events in 
>this format". 

When you indicate the stuff you support in an OPTIONS response, I agree
it's "I can support this format".

But, if I in an INVITE response indicate support of a certain content
type I think that can also be seen as "When/if you are going to send me
X, please use this format".

>This is why we didn't go with the INFO method for DTMF. Since DTMF
would already be arriving at the UAS in 
>the form of 2833, the primary entities needing signaling-based
solutions are applications not on the media 
>path.

In most cases, yes, and my intention is not to promote people to use a
signaling-based solution instead of 2833. We could even make very strong
recommendations to always use 2833, when possible.

(However, in theory you could also have a signalling gateway, without
access to media, which sends the DTMFs using a signaling-based solution
since the protocol "on the other side" also does it.)

>But, we only want to send such signaling when there is an application
that is interested. This is exactly what 
>SUB/NOT is for - to ask for notifications of events.

The application may not necessarily know that I am going to send him
anything, and then it will have to ask "just in case".
 
>Furthermore, by using a separate dialog we could have the 
>notifications sent just to the interested application server, 
>so that the proxies on the path of the original dialog don't 
>need to see INFO messages flying by if they weren't interested.

That is an advantage, yes.

...eventhough I wonder how many proxies you actually would be able to
"skip" in real life scenarios. You may have routes established using
Path and/or Service-Route, for example.

>Consider further the other UA in the call. If it supports 
>both DTMF-over-INFO and RFC2833, what does it mean to get 
>both? If you generate DTMF on receipt of INFO, you stand a 
>real risk of generating double DTMF which is very bad.

You have exactly the same problem with SUB/NOT, if you BOTH subscribbe
to the DTMF event AND inciate support of 2833 in your SDP.

So, the solution is: you indicate EITHER support of 2833 OR the
signaling-based mechanism (whatever it is). 

>You avoid this problem if you only send DTMF over signaling when asked;
this way the thing getting the rfc2833 doesn't 
>ask for it.

I am not sure I understand that. Even if I use SUB/NOT I don't know
exactly when the DTMF event will come (if it will come), and I could
still also receive 2833 while the subscription is active.

>Using SUB/NOT also allowed the application to customize when it would
get a notification. So for example an app that 
>needed each digit as it was pressed could get them immediately. One
only needing them after some
>pattern was matched could only get it then. This was also a perfect
match for SUB/NOT.

Well, obviously people using INFO do not need the subscriber to be able
to "push" the digitmap. But, if I am asked to give 4 digits I may
implement a user interface which allows me to enter the digits before
they are sent in a single INFO (assuming the INFO based mechanism allows
it).

>So there are serious technical limitations to carrying DTMF 
>over INFO. 
>Sure, it can 'work', but it is very suboptimal and can cause problems.
> 
>As for a general INFO usage, my opinion is unchanged on this. 
>We have numerous methods for extending SIP by adding new 
>headers and methods and event packages. Can you demonstrate a 
>use case where one of these extensibility mechanisms is not 
>sufficient for some use case?

>From a functional perspective I am pretty sure they are sufficient.

And, I am not saying that all INFO usages are good, and that you
couldn't use a better (both from a functionality and complexity
perspective) alternative.

Regards,

Christer


_______________________________________________
Sip mailing list  https://www1.ietf.org/mailman/listinfo/sip
This list is for NEW development of the core SIP Protocol
Use [EMAIL PROTECTED] for questions on current sip
Use [EMAIL PROTECTED] for new developments on the application of sip

Reply via email to